By Dr. David Clark

11/20/23 Inhibitor development in hemophilia is still a mystery.  There is a lot that we don’t know.  One question that has come up over the years is whether the product used to treat the hemophilia patient affects inhibitor development.  The answer has never been clear.  Recently a large group of European and Canadian researchers decided to re-address that question.  They looked at previously-untreated patients (PUPs – usually kids) with either hemophilia A or B at 56 European treatment centers and 23 Canadian centers.  They found 312 subjects out of 1219 total hemophilia A severe PUPs who developed inhibitors, about 26%.  For the As, inhibitor development was lower on plasma-derived factor VIII products; highest on standard half-life (SHL) products and intermediate on extended half-life (EHL) products.  Inhibitor development rates also varied among the various SHL and EHL products.

For hemophilia B, the situation appears to be different.  First, overall Bs had much lower inhibitor development rates, an average of 8% (14 of 173 study subjects).  And, because there were many fewer subjects, the statistics were not able to show any significant difference among products.  The results were that 11% (CI: 3-25%) developed an inhibitor on plasma-derived products, 8% (CI: 3-15%) on SHL products and 7% (CI: 1-22%) on EHL products.

Now, you might be thinking that’s the answer:  plasma-derived is worse than SHL, which is worse than EHL.  You would be wrong, and that’s why I included the confidence intervals (CI) with the numbers.  In fact, this study doesn’t show any difference among the products.

The confidence intervals (CI) show the statistical ranges in which we believe that the result lies.  For instance, with plasma-derived products the result is 11%, but we can only say with confidence that the actual number, if we included all possible subjects, is between 3% and 25%.  To reduce the size of the confidence interval would require more subjects, but subjects with hemophilia B and an inhibitor are harder to find.  Since the confidence intervals all overlap, we can’t say that the results are different.

Statistics are extremely important in science and medicine.  Although the media hardly ever publish the statistics for the public to see, scientific publications usually do.  Otherwise, we could be easily misled.  Using the above example, for instance, I could say that 11% of hemophilia B PUPs develop inhibitors on plasma-derived products.  My friendly-competitor in the next lab, however, could completely accurately tell me I was wrong – I only showed that the inhibitor rate is somewhere between 3% and 25%.  That’s quite a difference, and I should be careful about staking my argument on the fact that I think the result is exactly 11%.  Statistics show that 3%, 11%, 25% or anything in between are equally valid results from my data.  (There goes my Nobel prize!)

Normally, we don’t show you the statistical results, but I’m showing them to you here, so you can get a better idea of how science actually works.  We always look at the statistics before reporting findings in this newsletter, even if we don’t show them.

You don’t just do an experiment and get a result and that’s that.  Usually, you do a number of experiments so you can get a statistical idea of how good your answer is.  In medicine especially, we have to have good confidence in our results.  We don’t just test one subject and assume that applies to all other patients.  We have to test a number of subjects and average the results to get a good idea of what’s actually going on.  The more subjects we test; the better our answer is.

So, the next time that your neighbor boasts that he’s getting 50 mpg from his new car, ask him what the confidence limits are on that number – maybe he had a tail wind?  [Fischer K et al., Res Pract Thromb Haemost, online ahead of print 11/20/23]